Transcript Slide 1

Statistical Impact/Influence on
Regulatory
Brent Harrington
Director
Pharm Sci & PGS Statistics
Pfizer Inc
2014 Midwest Biopharmaceutical Statistics Workshop
May 19-21, 2014
7/17/2015 4:55 PM
1
Acknowledgements








Kimber Barnett
James Bergum5
Todd L. Cecil2
Tom Garcia
Tim Graul
Oliver Grosche3
Melissa Hanna-Brown
Jeff Hofer4
1.
2.
3.
4.
5.





Phil Nethercote1
John Peterson1
Kim Vukovinsky
Chad Wolfe4
Loren Wrisley
GlaxoSmithKline
United States Pharmacopeia
Novartis
Eli Lilly
Consultant
7/17/2015 4:55 PM
2
Outline
• A Few Successful Applications of Statistical
Methods and Thinking
• A 20+ year reflection
• Pharmaceutical Development in the 21st
Century
• Successful Applications of Statistical Methods
in the 21st Century
• Opportunities
7/17/2015 4:55 PM
3
Success Story … 20th Century
Analysis of Stability Results – Confidence on the Mean
utilizing ANCOVA
1987 Guideline for Submitting Documentation For The Stability of Human
Drugs and Biologics
7/17/2015 4:55 PM
4
Success Story … 20th Century
Multiple Criteria for Standards; USP<711>, USP<905>
Table 1. USP<711> Acceptance Table 1
Stage
S1
Number
Tested
6
S2
12
S3
24
Acceptance Criteria
Each unit is not less than
Q+5%
Average of 12 units (S1+S2)
is  Q, and no unit is <Q-15%
Average of 24 units
(S1+S2+S3) is  Q, not more
than 2 units is < Q-15%, and
no unit is <Q–25%5
7/17/2015 4:55 PM
5
Success Story … 20th Century
Probability of Meeting Standards
95% Lower Bound
(X  Z*ULS/n, ULS)
Acceptance Region
(N=100)
Acceptance
Region
Acceptance
Region
(n=10)
(N=10)
(X,
(X,S)S)

Confidence
Interval
Confidence
Interval
(n=10)
(N=10)
7/17/2015 4:55 PM
6
Success Story … 20th Century
Fractional Factorial Application to Stability Study Designs
Examples provided in ICH Q1D
Strength
Batch
Cont.
Size
50 ct
50 mg
A
T1
100 ct
T2
500 ct
B
T2
C
75 mg
A
B
T2
T1
T1
T2
T1
T1
T2
C
T1
100 mg
A
B
T1
T2
T1
C
T2
T2
T2
T1
7
7/17/2015 4:55 PM
7
New Initiatives!! – Moving
Beyond the 20th Century?
•
Inititiative: FDA – Pharmaceutical Quality in the 21st
Century – 2003
• Encourage Application of science to better understand
processes
•
Guidance Documents
•
•
•
•
•
•
FDA – Blend Uniformity – 2003
FDA – Analytical Validation - 2014
ICH – Q1A(r2), Q1D, Q1E – 2003
ICH –Q2, Q6, Q8, Q9, Q10, Q11 – 2005-2012
USP<1010> - 2010
USP Stimuli to the Revision Process
•
Acceptable, Equivalent, or Better… - 2009
•
Performance Based Monographs - 2009
7/17/2015 4:55 PM
8
Risk Assessment + DoE Strategy
Standard & Sample Preparation
Chromatography (Method)
Result
The level of effort, formality and documentation of the
quality risk management process should be
commensurate with the level of risk (ICH Q9)
7/17/2015 4:55 PM
9
Design of Experiments
Method Development Experiments – Define the ranges that
provide control of the measurement system
DoE Strategy - sequential exploration and modeling dependent upon the knowledge of the system, and the costs/resources
available.
7/17/2015 4:55 PM
10
Contour Plots of Two Responses &
Two PP’s
Impurity1 (%)
1.00
Impurity2 (%)
1.00
0.50
0.50
C
C
0.03
0.00
0.07
3
0.02
0.00
0.05
0.07
0.09
0.11
3
0.10
-0.50
-0.50
0.14
0.17
-1.00
-1.00
-1.00
-0.50
0.00
B
0.50
1.00
-1.00
-0.50
0.00
0.50
1.00
B
11
7/17/2015 4:55 PM
11
Overlay Plot of Two Responses vs. Two
PP’s
• Easy to implement.
• Lends to “Edge of
Failure” Terminology
• “EOF” is misleading.
12
7/17/2015 4:55 PM
12
Overlay Plot of Two Responses vs.
Two PP’s
~50% Prob
< 50% Prob
~50% Prob
• The probability of
simultaneously passing the
specifications varies within
in the yellow region
• “Boundary” provides no
greater than 50% probability
of passing
• The combined probabilities
decrease in areas of
intersecting requirements
13
7/17/2015 4:55 PM
13
Success Story … Prospective Process
Reliability Estimate (PPRE)
7/17/2015 4:55 PM
14
Issues?
• Reality – Pharmaceutical Development is highly
regulated
• Real or perceived regulatory hurdles cause
Compliance-driven rather than Process
Understanding-driven development
– No clear definition of relevant measure exists
– What measure (reportable value) assessed
against specifications?
– No incentive for providing additional testing to
increase confidence in scientifically risk-based
decisions
7/17/2015 4:55 PM
15
Regulatory Criteria
• ICH Q6A/B (Specifications):
• Acceptance Criteria are limits that each individual reportable
value is expected to meet
• ICH Q1E (Evaluation of Stability Data):
• The 95% confidence limit(s) for the batch mean must
remain within Acceptance Criteria to justify the product
expiry dating.
• The Acceptance Criteria are assumed to be known with
dating justification as the goal.
• USP:
• Not statistical tests, but standards
• Any official article is expected to meet compendial
standards if tested
• Inference only applies to the tested units
7/17/2015 4:55 PM
16
Regulatory Criteria
FDA Out of Specification (OOS) Guidance
• “In cases where a series of assay results (to produce a single reportable
result) are required by the test procedure and some of the individual
results are OOS, some are within specification, and all are within the
known variability of the method, the passing results are no more likely to
represent the true value for the sample than the OOS results.
• For this reason,
• This approach is consistent with the principle outlined in the USP
General Notices that an official article shall comply with the compendia
standard any time a compendia test is applied. Thus,
.”
Chad Wolfe and Jeffrey Hofer, Eli Lilly
177/17/2015 4:55 PM
17
Regulatory Criteria
Interpretation of Regulatory Acceptance Criteria
Topic
ICH
Specifications Overall batch mean (Q1)
USP
FDA OOS
Guidance
Reportable Value
Each replicate in a
Reportable Value
Reportable Value (Q6)
Individual
dosage units?
Reportable
values?
All of the above interpretations are
not aligned
Batch
Mean?
7/17/2015 4:55 PM
18
Regulatory Criteria
• Specifications are arrived at via negotiation
• Limited manufacturing experience at time of negotiation
• Rarely clinically driven but typically performance driven –
this is the case for drug product potency, but not
necessarily impurities
• Established based upon reportable values
• Often minimally wider than capability (3-sigma) limits
• Leads to low capability by definition
7/17/2015 4:55 PM
19
Process Capability … the
reality
Negotiation with
regulators often
ends up with
specs similar to
the middle
picture
Cp < 1
Cp = 1
On the edge of
capability for
reportable values
Cp > 1
7/17/2015 4:55 PM
20
The Measurement System
• Complex
• Destructive
• Atypical results rare but inevitable
• Lack of clarity on acceptable investigation practices
• Often most observed variation is measurement
• Little true process variability
• Little discrimination ability within specification range
7/17/2015 4:55 PM
21
Graphical
GraphicalRepresentation
Representationof
of
Traditional
ATPMethod
MethodValidation
ValidationCriteria
Criteria
22
The procedure
• Accuracy
andmust be able
to
Precision with
accurately quantify the Drug
Independent
product over a range of 98%
limits/criterion
to 102% of the nominal
• concentration
No trade-offwith accuracy,
and precision such that
between
increasing
measurements
fall within ±
method
biastrue
and/or
2.0% of the
value at
least a 95% probability.
variability
Criteria:
• Bias of NMT 2.0% &
• Variability of NMT 1.25%
7/17/2015 4:55 PM
22
Success – ATP verification
Appropriate Sample Size to Assess Acceptance Criteria
Developing “Equivalent or Better” Methods
Verification of meeting
ATP criterion is
illustrated via
simultaneous conf
interval on accuracy
and precision
estimates from data
realized under
experimental
conditions
*E2709 – 10: Demonstrating Confidence in Complying with Acceptance Procedures
7/17/2015 4:55 PM
23
Success – Potency Assay Sample Strategy
Preparation is 20% and Dosage Unit is 80%of Total Variability
7/17/2015 4:55 PM
24
Opportunities
• Define reportable result as reliable estimation of
batch parameter
• Summary measures such as the mean provide
greater information about the attribute
• Differentiate between acceptance criteria (limits)
and process control limits
• Acceptance criteria should ensure safety and
efficacy performance
• Data-driven signals define what the process is
capable of delivering
• Recognize the value of additional testing
• Link sample size and acceptance criteria to
manage risk
7/17/2015 4:55 PM
25
UDU Acceptance
Link Sample Size to Acceptance Criteria
Use OC of test
to determine
discrimination
7/17/2015 4:55 PM
26
Are there other Methods?
Relevance?
Probability of Meeting Standards
95% Lower Bound
(X  Z*ULS/n, ULS)
Acceptance Region
(N=100)
Acceptance
Region
Acceptance
Region
(n=10)
(N=10)
(X,
(X,S)S)

Confidence
Interval
Confidence
Interval
(n=10)
(N=10)
7/17/2015 4:55 PM
27
Path Forward
• Continue to develop appropriate experimental plans
and methods that promote risk-based decision
making
• Engage R&D, Compendia and Regulatory
authorities to work together in constructive
dialogue concerning measurements and inferences
made from those measurements
• Systems should not discourage the gathering of
greater amounts of data, and appropriate statistical
techniques must be used when needed.
7/17/2015 4:55 PM
28
Thank you, for your attention!
29
Abbreviated Reference List:
1. Pharmacopeial Forum, Vol. 35(3) [May-June 2009], Stimuli to the Revision Process;
Performance-based Monographs, 765-771
2. Fanali, S., Haddad, P.R., et. al. (2013). Liquid Chromatography Applications. Waltham, MA:
Elsevier
3. EURACHEM Working Group. (2000). Quantifying Uncertainty in Analytical Measurement.
4. Draft paper A Simple Parametric Bootstrap Approach to Computing Design Space
Probabilities by John Peterson of GlaxoSmithKine
5. Reid, George; et al; Analytical Quality by Design (AQbD) in Pharmaceutical Development.
American Pharmaceutical Review. Vol. 16, issue 5.
6. 1987 FDA Guideline: “Guideline for Submitting Documentation for the Stability of Human
Drugs and Biologics”
7. ICH Quality Guidance: Q1A(r22), Q1D, Q1E, Q8, Q9, Q10, Q11
8. USP General Chapters: <1010>, <724>, <905>
9. Chow, Liu (1995). Statistical Design and Analysis in Pharmaceutical Science. Dekker. NY.
7/17/2015 4:55 PM
29